Section: Systems and Control
Tutor: GARATTI SIMONE
Advisor: PANZANI GIULIO Major Research topic
:Sensing And Dynamics Control For Assisted And Automated Drive in MotorcyclesAbstract:
Autonomous driving represents the biggest challenge towards the future of automotive industry. On the long run, the benefits possibly deriving from the introduction of such paradigm are numerous and relevant: from the improvement of vehicle's safety, resulting in a significant reduction of road accidents related casualties, to the enhancement of users comfort, cancelling downtime and stress associated with driving. Even though we are still distant from reaching the goal of full automation, the advantages related to the extensive research along these lines are already noticeable and substantial. In this sense, it is worth mentioning Advanced Driver Assistance Systems (ADAS), that are electronic systems designed to help the driver when facing many challenging driving situations. ADAS constitute a clear example of technological advance introduced as a direct consequence of research aiming towards unmanned vehicles, representing a first step in the introduction of on-board automation.
What stated above is true in general regarding four-wheeled vehicles. Considering the motorcycle field instead, the development of such technologies is still at an embryonic stage, despite the possible benefits that could derive from their large scale adoption. In fact, even if an autonomous two-wheeled vehicle could be hardly conceived to carry any passenger, its intrinsic agility and inclination to avoid remaining stuck in traffic make it suitable to be employed in deliveries and similar applications. Moreover, as for the four-wheeled vehicle case, the research in such direction could speed up the development of rider assistance systems, enhancing the safety of a vulnerable road users category, subject to a high risk of serious injuries in the event of a crash.
For these reasons, this Dissertation deals with the design and implementation of sensing and control algorithms oriented towards autonomous and assisted drive of single-track vehicles. Given the broadness of the general topic, that encompasses a wide number of different sub-problems, the Work is focused on a number of systems - listed in the following - considered particularly relevant for the field of application.
Firstly, the problem of designing a two-wheeled vehicles oriented lane detection algorithm based on a single calibrated camera is assessed, robust with respect to roll angle dynamics. The system aims at recognizing the current driving lane, thus retrieving a key information to plan the motorcycle motion and reconstructing the driving environment. The output of the lane detection stage is also exploited to develop a braking manoeuvre detection algorithm for vehicles preceding the motorcycle, conceived to rapidly react to hazards deriving from possible frontal crashes.
Then, an autonomous parking strategy for a three-wheeled scooter based on a low resolution Lidar is presented, designed to help the user to drive the vehicle inside very narrow parking spots delimited by two physical obstacles, a particular use case that would make it difficult and uncomfortable for the rider to get off at the end of the parking manoeuvre.
At last, the development of a path tracking algorithm for a two-wheeled scooter is discussed, with the objective of correctly following a reference trajectory, an issue of primary importance in the autonomous driving context.
While the presented algorithms are mainly designed as different subsystems to be integrated in the two-wheeled oriented autonomous drive framework, nevertheless they could be easily adapted to function as standalone rider assistance systems aimed to improve comfort and safety of the final user.